Deep Tracking by Tagging


Short name:

DpTrack_16

Benchmark:

Description:

Public DPM detections are used and the tracking is achieved by assigning each person a different tag on an 8 frame video clips. By sliding 4 frame at a time and using the overlapping information tracking is achieved through out the video.

Hardware:

3.0GHz,1_Core_and_Titan_X_Pascal

Detector:

Public

Processing:

Batch

Last submitted:

May 08, 2019 (3 months ago)

Published:

July 12, 2019 at 22:21:19 CET

Submissions:

3

Open source:

No

Project page / code:

n/a

Reference:

Anonymous submission

Benchmark performance:

MOTAMOTPFAFMTMLFPFNID Sw.FragSpecificationsDetector
56.575.31.422.7 % 30.8 % 8,14968,8402,3872,9303.0GHz,1_Core_and_Titan_X_PascalPublic
IDF1ID PrecisionID Recall
45.056.337.5

Detailed performance:

Sequence MOTA IDF1 MOTP FAF GT MT ML FP FN ID Sw Frag
MOT16-0141.634.473.01.82334.8 % 17.4 % 8122,84775151
MOT16-0370.350.175.81.814845.9 % 10.8 % 2,76827,7285861,312
MOT16-0649.245.873.00.722124.9 % 32.1 % 8584,540464237
MOT16-0744.838.574.31.95420.4 % 24.1 % 9647,847197304
MOT16-0832.130.577.01.26312.7 % 36.5 % 72310,471169249
MOT16-1244.655.877.60.58619.8 % 43.0 % 4804,0714562
MOT16-1425.727.771.62.11643.0 % 42.7 % 1,54411,336851615

Raw data:

n/a


DpTrack_16